fix(client): sleep between polls in the blocking poll helper

[cplieger]

Two small defects in the execute-then-poll helpers, found while working out why an Action's second execute on one resource failed with Resource is busy. (It was my bug: I should have used execute_and_poll. These two turned up on the way.)

1. The blocking poll helper busy-loops

poll_update_until_complete under #[cfg(feature = "blocking")] has no delay in its loop, so it re-issues GetUpdate as fast as the network allows for the whole duration of the execution. For a Stack deploy that is minutes of hot looping against Core.

The async variant already does tokio::time::sleep(Duration::from_millis(500)). This adds the matching std::thread::sleep, same interval. Duration is already imported.

2. A doc comment points at a function that does not exist

client/core/ts/src/lib.ts, on execute:

To have the call only return when the task finishes, use [execute_and_poll_until_complete].

The exported function is execute_and_poll. The surrounding note is genuinely useful, which is why the dangling name is worth fixing: it is the pointer a reader follows when they hit exactly the problem the note describes.

Verification

cargo check -p komodo_client --features blocking clean, cargo fmt leaves it untouched. The TS change is a comment.